Bridging Modalities: Joint Synthesis and Registration Framework for Aligning Diffusion MRI with T1-Weighted Images

TL;DR

This work tackles the difficulty of cross-modal registration between diffusion MRI and T1-weighted images by transforming the problem into unimodal registration through an unsupervised joint synthesis–registration framework. A Brain-ID-based synthesis module creates a T1w-like image $T1w′$ from the dMRI $b0$, and a UNet-based registration module aligns $T1w′$ to $T1w$ using a Spatial Transformer, with a joint loss that couples similarity terms and a deformation smoothness term: $L_{total} = L_{sim}(I_f, I_m \circ \phi) + \lambda_1 L_{sim}(I_f, I_s \circ \phi) + \lambda_2 L_{smooth}(\phi)$. The method is evaluated on two public datasets (HCP and PPMI) against multiple baselines and ablations, consistently achieving superior Dice scores and lower ASD, indicating improved anatomical fidelity and deformation stability. The approach is unsupervised and end-to-end trainable, enabling iterative refinement of synthesis to guide registration and offering a robust pipeline for downstream diffusion analyses.

Abstract

Multimodal image registration between diffusion MRI (dMRI) and T1-weighted (T1w) MRI images is a critical step for aligning diffusion-weighted imaging (DWI) data with structural anatomical space. Traditional registration methods often struggle to ensure accuracy due to the large intensity differences between diffusion data and high-resolution anatomical structures. This paper proposes an unsupervised registration framework based on a generative registration network, which transforms the original multimodal registration problem between b0 and T1w images into a unimodal registration task between a generated image and the real T1w image. This effectively reduces the complexity of cross-modal registration. The framework first employs an image synthesis model to generate images with T1w-like contrast, and then learns a deformation field from the generated image to the fixed T1w image. The registration network jointly optimizes local structural similarity and cross-modal statistical dependency to improve deformation estimation accuracy. Experiments conducted on two independent datasets demonstrate that the proposed method outperforms several state-of-the-art approaches in multimodal registration tasks.

Figures (3)

• Figure 1: Framework of the proposed method.
• Figure 2: Warped images (row 1, columns 2–6) and the corresponding instance deformation fields $\phi$ (row 2, columns 2–6) after registering the moving images to the fixed images (column 1). Boundaries of the ventricles, thalamus, and hippocampus are overlaid to illustrate the plausibility of the deformations produced by our method.
• Figure 3: Visual comparison of registration methods on two datasets. Each row shows the warped images alongside the original and fixed images, with red boxes highlighting fine structural details.